Management of Bus Operations and Technology

Transcription

1 Management of Bus Operations and Technology Nigel H.M. Wilson June 10 th,2008

2 OUTLINE Applications of Automated Data Collection Systems Scheduling Connection Protection Service Quality Monitoring Travel Pattern Inference Travel Behavior Ongoing Bus Operations Research Scheduling Limited Stop / X Service Design Operations Management and Control Interchange Analysis 2

3 MIT Research Focus Get better information from Automated Data Collection Systems AFC: Automatic Fare Collection Systems AVL: Automatic Vehicle Location Systems APC: Automatic Passenger Counting Systems To support key agency functions: Service and Operations Planning Operations Management and Control Customer Information Performance Measurement and Monitoring 3

4 Scheduling Application Context: Rail scheduling and operations control Data: Focus: Train tracking data Analysis of train dwell times, headways, schedule adherence, time-space diagrams, animation playback capability Application: MBTA Red Line Recommendations: Schedule adjustments, branch offsets adjustments, tighter terminal departure discipline Key researchers: Haris Koutsopoulos, Matt Dixon, Zhigao Wang (2006) 4

5 Connection Protection Application Context: Transfers from rail to bus Data: Focus: Train tracking data Develop improved but simple dispatching strategy for buses based on bus holding lights linked to impending train arrival after long gap Application: MBTA Red Line Alewife Station Recommendations: Implementation would reduce transfer wait time by 25% and greatly reduce "near misses" Key researchers: Drew Desautels (2006) 5

6 Service Quality Monitoring Application Context: Rail operations Data: Focus: Smart card station entry and exit times Develop measures of service reliability from customer's perspective Application: London Underground service times using Oyster time data Recommendations: Use of reliability buffer as an additional measure of service delivery alongside mean measure (Journey Time Metric) Key researchers: Joanne Chan (2007), David Uniman (in progress) 6

7 Context: Data: Focus: Travel Pattern Inference Application Application: Any public transport network -- bus, rail, or combined AFC transactions (smart card or magnetic stripe) and AVL data Estimation of customer origin-destination travel patterns CTA Rail network London Underground network CTA Bus network Recommendations: A practical method for estimating travel patterns with virtually no additional cost beyond existing ADCs Key researchers: Jinhua Zhao (2004), Alex Cui (2006), Fabio Gordillo (2006), Joanne Chan (2007) 7

8 Basic Idea Each AFC record includes: AFC card ID transaction type transaction time transaction location: rail station or bus route TS 1 A (loc A, time A ) B (loc B, time B ) TS 2 TS 3 C (loc C, time C ) D (loc D, time D ) The destination of many trip segments (TS) is also the origin of the following trip segment. Note: 1) each bus boarding requires a new AFC transaction: TS bus represents an unlinked bus trip 2) rail-rail transfers do not require a new AFC transaction: TS rail represents a path on the rail network 8

9 Context: Data: Focus: Application: Travel Behavior Application Any public transport network -- bus, rail, or combined Registered smart card address (exact or approximate), smart card transactions, and AVL data Infer public transport access behavior and modal preferences CTA bus and rail network Recommendations: This represents a valuable way to monitor behavior and modal preferences as the system changes over time Key researchers: Mariko Utsunomiya (2005), Saumya Gupta (2006) 9

10 Path Choice Analysis: Sample Users Multiple rail and bus routes serving the loop High quality express bus service and rail service 10

11 Ongoing Bus Operations Research Scheduling Limited Stop / X Service Design Operations Management and Control Interchange Analysis Bus Route Simulation Model Development Chicago Loop Congestion Analysis 11

12 Bus Scheduling Bus service reliability is a chronic concern of customers and agencies Traditionally schedules are developed using rules of thumb -- not a real problem because data has been very poor on actual running times Now with AVL data we can evaluate alternative scheduling methods Goal is to develop schedules which: result in reliable service don't increase running time too much don't cost too much Key researchers: Grace Fattouche (2007), Clara Yuan (in progress) 12

13 Problem Statement On high-frequency service, i.e. headways minutes: Develop a timepoint level schedule which minimizes the total weighted time cost to customers Subject to: At least 90% of buses should be ready to start their next trip on time Operators should have at least 5 minutes of recovery time at one end of the route 13

14 Model Evaluates the cost for the waiting passengers, onboard passengers and CTA of a proposed schedule Assumes schedule-based holding strategy (i.e. operators do not depart time point early) Inputs: AVL and APC data, headway, segment running times and number of buses in operations Outputs: costs for the waiting passengers, onboard passengers and CTA 14

15 Application to CTA Route 95E CTA Key route Runs 5 miles East-West on 93rd St. and 95th St. Five time points Connects with the southern Red Line terminal, and two Metra stations High frequency: every 10 mins between 6:00 and 18:00 15

16 Percentage Change in Passenger Excess Time 100% 80% 60% 40% 20% 0% -20% -40% -60% -80% Current Schedule + Schedule-based holding strategy Traditional Approach (50th percentile) Traditional Approach (65th percentile) Generalized cost minimization approach % difference in excess waiting time % difference in excess in-vehicle TT % difference in excess pax-min % difference in excess weighted pax-min 16

17 Sensitivity Analysis The generalized cost minimization schedule is sensitive to: Ratio of waiting passengers to through passengers Location of the segment on the route Ratio of waiting passengers on later segments to through passengers Route Length 17

18 Limited Stop / X Service Design Many high-frequency bus routes in major metropolitan areas are long and have many stops, resulting in: long travel times poor reliability unattractive for long journeys Limited Stop / X Services are overlay routes with far fewer stops, which: provide alternative service mix, which attracts different markets, e.g. longer journeys reduce the interaction between buses, i.e., less bunching One step towards BRT Key researchers: Stacey Schwarcz (2004), Harvey Scorcia (in progress) 18

19 Problem Statement Establish guidelines for the addition of limited stop service Develop a model to help in design and evaluation of these services Key elements in limited stop service design: Reduction in # of stops Running time savings Headway split between local and limited stop service Resources: unchanged or increased 19

20 Model Components Key inputs: Demand by stop Bus running times Limited stops Frequency split Key processes: Stop choice for customers closest to local only stop Route choice for customers at combined stops Key outputs: Travel times Productivity Market Share Use of stops 20

21 CTA Route 9/X9 Description X9 Implemented in Summer X9 Length 20 miles Average Daily Ridership 35,000 Combined Headway 5 min. Split 50% Number of Stops Travel Time (mins): NB SB

22 Change in Platform Hours Route X9 required a modest increase in resources 22

23 Productivity Overall increase in productivity 23

24 Limited Stop Design Guidelines Stop spacing: CTA X services typically serve 25-30% of local stops Stop spacing is mile between X service stops Based on cumulative demand by stop function cumulative demand percent of total stops 24

25 Limited Stop Design Guidelines Running time reduction: Savings of at least 15% should be achievable Limited stop frequency share: Frequency on X routes must be greater than on local service Typically 60% of service should be on the X route 25

26 Operations Management and Control Objectives: Examine how real-time AVL system can support improved service reliability Application: CTA Route 20 AM peak period headway: 5 minutes Average weekday boardings: 24, mile, 60 minute eastbound trip Capacity and bus bunching issues Key researchers: Chris Pangilinan (2006) 26

27 Operations Management and Control Application: CTA Bus Tracker System MIT Student acted as controller for one week using Bus Tracker information Individual Buses 27

28 Results Headways leaving Austin (Terminal) 70% 60% 50% 40% 30% 20% 10% 0% 0 to 2 2 to 4 4 to 6 6 to 8 > 8 Headway at Austin (min) Previous Week Experiment Period January 15, 2008 Chicago Transit Authority / MIT 28 28

29 Results Headways 3.5 miles east of terminal 70% 60% 50% 40% 30% 20% 10% 0% 0 to 2 2 to 4 4 to 6 6 to 8 > 8 Headway Ratio at Kedzie Previous Week Experiment Period January 15, 2008 Chicago Transit Authority / MIT 29 29

30 Results: Excess Wait Time Austin Cicero Pulaski Kedzie Ashland Halsted Timepoint Previous Week Experiment Period % of Scheduled Wait Time Austin Cicero Pulaski Kedzie Ashland Halsted Previous Week 23% 38% 50% 58% 55% 77% Experiment Period 9% 15% 20% 23% 22% 31% 30 30

31 Results in Perspective To achieve the same reduction in excess passenger waiting time without improving reliability: 6 more buses would have to be added to the current 24 during the AM peak January 15, 2008 Chicago Transit Authority / MIT 31 31

32 Summary of Results Real-time AVL and supervision strategies contributed to: Lower rate of bus bunching Lower rate of long headways Downstream headways benefited from terminal control Reduction in excess wait time Dwell time, traffic, and other factors will take its toll on reliability, regardless of supervision January 15, 2008 Chicago Transit Authority / MIT 32 32

33 Interchange Analysis Research Questions: Can Oyster data be used to help improve the public transport network in London by focusing on bus passenger interchange behaviour? Key contribution: Methodology for exploring passenger interchange behaviour in London using Oyster card data Key researchers: Catherine Seaborn (in progress) 33

34 Journey Segments Per Passenger 40.0% 35.0% 34.9% Percentage of Total Passengers 30.0% 25.0% 20.0% 15.0% 17.1% 14.7% 14.1% 10.0% 5.0% 7.0% 4.8% 0.0% 2.7% 1.8% 1.1% 0.7% 0.4% 0.3% 0.1% 0.2% 0.1% Number of Daily Journey Segments Source: 5% Oyster data for 2007 Period 2 (April 29 May 26) 34

35 Consecutive Journey Segments 800, ,442 Total Journey Segment Pairs 700, , , , , , , , ,806 First to Second Journey Segments All Journey Segments 100, % 35% 56% 25% Bus-Underground Underground-Bus Bus-Bus Underground-Underground Journey Segment Mode Sequence Source: 100% Oyster data for Wednesday, November 14,

36 Weekday Journey Segment Patterns Top 10 patterns shown Total patterns: 15,802 Mode 1 Mode 2 Mode 3 Mode 4 Mode 5 Mode 6 Passengers Share Cumulative Share U U 416, % 16.3% B B 401, % 32.0% B 266, % 42.4% B B B 150, % 48.3% B B B B 144, % 54.0% U 125, % 58.9% B U U B 77, % 61.9% B B B B B 72, % 64.8% U U U 65, % 67.3% B B B B B B 50, % 69.3% Source: 100% Oyster data for Wednesday, November 14,

37 Underground-Underground Journey Segment Entry Times 40,000 35,000 30,000 Total Entries 25,000 20,000 15,000 1st Undgrnd Journey Segment 2nd Undgrnd Journey Segment 10,000 5,000-4:30 5:30 6:30 7:30 8:30 9:30 10:30 11:30 12:30 13:30 14:30 15:30 16:30 17:30 18:30 19:30 20:30 21:30 22:30 23:30 0:30 1:30 2:30 3:30 Time Period (15 minutes) Source: 100% Oyster data for Wednesday, November 14,

38 Bus-Bus Journey Segment Boarding Times 18,000 16,000 14,000 12,000 10,000 8,000 1st Bus Journey Segment 2nd Bus Journey Segment 6,000 4,000 2, :30 5:30 6:30 7:30 8:30 9:30 10:30 11:30 12:30 13:30 14:30 15:30 16:30 17:30 18:30 19:30 20:30 21:30 22:30 23:30 0:30 1:30 2:30 3:30 Total Boardings Time Interval (15 minutes) Source: 100% Oyster data for Wednesday, November 14, 2007

39 Underground Station Exit to Bus Boarding: Time Difference at Highest Exit Volume Stations 39

40 Bus Boarding to Underground Station Entry: Potential Transfers are Large Share of Entries 40